Seldon

Seldon (Seldon Technologies Limited, also referenced as Seldon.io) is a British software company headquartered in London that builds machine learning operations (MLOps) and LLMOps tooling for deploying, managing, monitoring and scaling AI and machine learning models in production. Its approach is cloud-agnostic and Kubernetes-native, intended to let enterprises run inference workloads on-premise or across public clouds without vendor lock-in.

The product line spans open-source and commercial components. Seldon Core 2 is the modular, data-centric deployment framework, orchestrating models as Kubernetes microservices and connecting them into real-time pipelines using Kafka for data streaming between components; it supports REST and gRPC, autoscaling, multi-model serving, overcommit with LRU memory swapping, and experiments such as A/B tests, canary and shadow deployments. MLServer is a lightweight multi-framework inference server compatible with the Open Inference Protocol. Alibi Detect provides outlier, adversarial and drift detection across tabular, image, text and time-series data, while Alibi Explain offers local, global, black-box and white-box explanation methods for classification and regression. An LLM Module adds GenAI deployment with prompt orchestration, RAG and memory design patterns, observability and configurable guardrails; a Model Performance Metrics (MPM) module covers production model quality monitoring; Seldon Core+ is a supported accelerator program; and enterprise tiers add authentication, audit trails and team controls for regulated industries. An earlier enterprise product, Seldon Deploy, automated ML engineering tasks on top of Seldon Core.

The platform integrates with Prometheus, Grafana, Kafka, Jaeger, Elasticsearch, Triton Inference Server, MLflow, Weights & Biases, Istio, Envoy, ArgoCD, Flux, Hugging Face, LangSmith, Databricks and rclone storage backends, and is tested on AWS EKS, Azure AKS, Google GKE, Alicloud, Digital Ocean and OpenShift.

Founding story

Seldon Technologies Limited was incorporated on 28 August 2014 in London by Alex Housley (CEO; later CPTO) and Clive Cox (CTO), entrepreneurs who had previously built mobile and data companies. The founding aim was cloud-agnostic tooling to make enterprise machine learning models deployable and valuable in real-world applications, with an initial focus on the deployment stage of the ML pipeline.

Business model

Seldon combines freely available open-source projects — Seldon Core 2, MLServer, Alibi Detect and Alibi Explain — with paid enterprise offerings including Core+, the LLM Module, the MPM module and an enterprise governance tier. The company describes its pricing as modular, so customers pay for the components they use. Management has said the open-source distribution channel allows proprietary solutions to be distributed more efficiently and cost-effectively.

Commercial licensing of enterprise products and modules layered on top of the open-source core, sold to enterprises; Seldon Core 2 is distributed under the Business Source License.

Traction

Seldon Core had over 700,000 models deployed by late 2020 and more than one million by 2020 per Wikipedia, with the company reporting 38% month-on-month growth since September 2018 at the time of its Series A. It reported 400% year-on-year growth in open-source frameworks installed and running between the Series A and the March 2023 Series B, and its site cites over 2 million installs. Named customers include PayPal, Johnson & Johnson, Audi, Experian, Capital One, Verizon, IKEA and Volkswagen, with the website also listing Covea, AstraZeneca, GSK, Aselsan, Noda and Cambridge University. Users are reported to see productivity gains of up to 92% and an average 84% reduction in deployment time.

Latest developments

Seldon has been acquired by TrueFoundry, and the seldon.io homepage now presents the combined offering as a single cloud-agnostic foundation spanning real-time inference and agentic AI, with existing Seldon enterprise customers offered a migration path toward agentic and composite AI workloads. TrueFoundry is cited as having reduced enterprise deployment timelines by more than 50% for global Fortune 500 firms.